Neuron Explorer

Qwen3-0.6B

the word 'social' in set phrases

A neuron in layer 22 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the word 'social' in phrases like 'social system', 'social contract', 'social media' and 'social organization'. It is most active when the next word is likely to be a noun about society. Its negative side is weaker and appears on words like 'biological' and 'on'. Turning it up gave text that drifted toward community and way of life, which is only loosely related.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    a nation acknowledging the threat of climate change, and doing something concrete to help become part of the solution.⏎As Greenpeace researcher Lauri<|endoftext|>Capitalism is a social system based on the recognition of individual rights, including property rights, in which all

  2. #2 peak

    the word to later explore its significance in more depth. Ideology: “The dimension of social experience in which meanings and values are produced”. Furthermore, he continues exploring the first

  3. #3 peak

    But it did work. People got jobs, got houses, got some basic needs taken care of.⏎That was the world our ancestors lived in. That was the social contract before the New Deal. There was no Medicaid for the poor, no subsidized

  4. #4 peak

    critical to closing the digital divide, and essential for boosting economic growth. Internet access allows women to gain a voice, seek information, or expand their networks beyond existing social boundaries.⏎Yet, African governments are failing to spend funds collected specifically to connect the

  5. #5 peak

    industrialism are created on this cell of thinking. Especially Marxism, in which all interactions are reduced to defining the determination of the Matter or Economy in a society: social<|endoftext|>Touring the Milky Way’s a blast with this brand new 36

  6. #6 peak

    3 states. The researchers analyzed the language of the social media posts by lecturers, mother and father, and college students and located elevated

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

turned down a lot
Prompt

Here is something I have been thinking about lately: I have been in the school school and I have been in the school school school... . The school is in the middle of the school... . The school is in the school school... . The school is in the school... . The school

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    Currently, the International Union for Conservation of Nature maintains the Red Data Book. IUCN is the world’s most detailed inventory centre of the global conservation status of biological species. The International Union for Conservation of Nature (IUCN) was founded

  2. #2 peak

    solve particular architectural problems. Readers will explore:⏎The five formal properties of composition—number, geometry, proportion, hierarchy, and orientation⏎How the plan conveys the central

  3. #3 peak

    on the right of the female population, thereby indicating the sexes. Pyramids are of two types: One, based on absolute numbers of population and the second, on the proportion of population. The former shows the size and composition of population of a

  4. #4 peak

    laws of physics and chemistry (a mixture of chance and necessity or "natural law") and not by design. This is the naturalistic explanation. Darwinism (evolution based on the hypothesized natural selection of random mutations) is the mechanism that

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L20:635 weakly excites
  2. L17:1189 weakly suppresses
  3. L20:209 weakly suppresses
  4. L16:811 weakly suppresses
  5. L10:550 weakly excites
  6. L8:1498 weakly excites
  7. L20:403 weakly suppresses
  8. L5:1994 weakly suppresses
  9. L11:1470 weakly excites
  10. L14:930 weakly excites
  11. L7:456 weakly suppresses
  12. L16:2642 weakly suppresses
  13. L21:942 weakly excites
  14. L16:1781 weakly suppresses
  15. L11:1968 weakly suppresses
  16. L6:3047 weakly suppresses
  17. L0:1138 weakly excites
  18. L10:2577 weakly suppresses
  19. L19:2426 weakly suppresses
  20. L9:117 weakly suppresses
  21. L6:761 weakly excites
  22. L1:1855 weakly excites
  23. L17:1791 weakly suppresses
  24. L20:2139 weakly excites

Feeds later layers

  1. L24:2299 weakly suppresses
  2. L25:1478 weakly suppresses
  3. L24:1504 weakly excites
  4. L24:2179 weakly excites
  5. L25:1974 weakly suppresses
  6. L23:97 weakly excites
  7. L23:2850 weakly suppresses
  8. L25:2639 weakly suppresses
  9. L23:333 weakly suppresses
  10. L23:532 weakly excites 'and' or 'press' in how-to steps
  11. L26:1764 weakly suppresses
  12. L24:1647 weakly suppresses
  13. L27:288 weakly suppresses
  14. L24:1723 weakly excites
  15. L23:2118 weakly suppresses
  16. L25:2925 weakly excites
  17. L25:2739 weakly excites
  18. L24:1669 weakly excites
  19. L25:1966 weakly excites
  20. L27:1840 weakly suppresses
  21. L25:1713 weakly suppresses
  22. L24:2656 weakly suppresses
  23. L23:1590 weakly excites
  24. L26:1974 weakly suppresses

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

1 values between -18.31 and -16.803 values between -13.76 and -12.241 values between -12.24 and -10.738 values between -10.73 and -9.219 values between -9.21 and -7.6912 values between -7.69 and -6.1829 values between -6.18 and -4.6636 values between -4.66 and -3.14181 values between -3.14 and -1.621010 values between -1.62 and -0.112310 values between -0.11 and 1.41170 values between 1.41 and 2.9369 values between 2.93 and 4.4535 values between 4.45 and 5.9619 values between 5.96 and 7.484 values between 7.48 and 9.003 values between 9.00 and 10.515 values between 10.51 and 12.031 values between 34.79 and 36.311 values between 37.82 and 39.341 values between 40.86 and 42.38offnegativepositiveoften

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.

Promotes

  1. .Infof
  2. rzy
  3. 먼
  4. 实践
  5. aiser
  6. едак
  7. eterangan
  8. sr
  9. работать
  10. tails

Suppresses

  1. kart
  2. reno
  3. ソン
  4. br
  5. вал
  6. ⼼
  7. R
  8. 庭
  9. печат
  10. (br